CORRECTION

Correction: Lysyl oxidases regulate fibrillar collagen remodelling inidiopathic pulmonary fibrosis (doi: 10.1242/dmm.030114)Gavin Tjin, Eric S. White, Alen Faiz, Delphine Sicard, Daniel J. Tschumperlin, Annabelle Mahar,Eleanor P. W. Kable and Janette K. Burgess

An error was published in Dis. Model. Mech. (2017). 10, 1301-1312 (doi: 10.1242/dmm.030114).

A reference was incorrectly cited in the Results section (subsection ‘Inhibition of LO activity reduced TGF-β-induced collagen remodellingin collagen I hydrogels’). The corrected sentence and reference are below.

‘To confirm the role of LO enzymes in fibrillar collagen remodelling, 3D in vitro cell culture experiments using inhibitors of LO activitywere performed. β-APN is a non-selective inhibitor of LO activity, whereas Compound A is a specific inhibitor of LOXL2 activity, referredto as PXS-S2A in a recent publication by Chang et al. (2017).’

Chang, J., Lucas,M. C., Leonte, L. E., Garcia-Montolio, M., Singh, L. B., Findlay, A. D., Deodhar, M., Foot, J. S., Jarolimek,W., Timpson,P. et al. (2017). Pre-clinical evaluation of small molecule LOXL2 inhibitors in breast cancer.Oncotarget 8, 26066-26078. http://doi.org/10.18632/oncotarget.15257.

The authors apologise to readers for this error.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution andreproduction in any medium provided that the original work is properly attributed.

1545

© 2017. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2017) 10, 1545 doi:10.1242/dmm.033191

Disea

seModels&Mechan

isms

RESEARCH ARTICLE

Lysyl oxidases regulate fibrillar collagen remodelling in idiopathicpulmonary fibrosisGavin Tjin1,2,3,4,*, Eric S. White5, Alen Faiz6,7,8, Delphine Sicard9, Daniel J. Tschumperlin9, Annabelle Mahar10,Eleanor P. W. Kable3 and Janette K. Burgess1,2,7,8,11,*

ABSTRACTIdiopathic pulmonary fibrosis (IPF) is a progressive scarring disease ofthe lungwith feweffective therapeutic options. Structural remodelling ofthe extracellular matrix [i.e. collagen cross-linkingmediated by the lysyloxidase (LO) family of enzymes (LOX, LOXL1-4)] might contribute todisease pathogenesis and represent a therapeutic target. This studyaimed to further our understanding of the mechanisms by which LOinhibitors might improve lung fibrosis. Lung tissues from IPF and non-IPF subjects were examined for collagen structure (second harmonicgeneration imaging) and LO gene (microarray analysis) and protein(immunohistochemistry and western blotting) levels. Functional effects(collagen structure and tissue stiffness using atomic force microscopy)of LO inhibitors on collagen remodelling were examined in twomodels,collagen hydrogels and decellularized human lung matrices. LOXL1/LOXL2 gene expression and protein levels were increased in IPFversus non-IPF. Increased collagen fibril thickness in IPF versusnon-IPF lung tissues correlated with increased LOXL1/LOXL2, anddecreased LOX, protein expression. β-Aminoproprionitrile (β-APN;pan-LO inhibitor) but not Compound A (LOXL2-specific inhibitor)interfered with transforming growth factor-β-induced collagenremodelling in both models. The β-APN treatment group was testedfurther, and β-APN was found to interfere with stiffening in thedecellularized matrix model. LOXL1 activity might drive collagenremodelling in IPF lungs. The interrelationship between collagenstructural remodelling and LOs is disrupted in IPF lungs. Inhibition of

LO activity alleviates fibrosis by limiting fibrillar collagen cross-linking,thereby potentially impeding the formation of a pathologicalmicroenvironment in IPF.

KEY WORDS: Idiopathic pulmonary fibrosis, Extracellular matrix,Collagen, Lysyl oxidase, Second harmonic generation

INTRODUCTIONIdiopathic pulmonary fibrosis (IPF) is a specific form of progressivefibrosing lung disease of unknown cause that leads to chronicrespiratory failure and death (Raghu et al., 2011), with a mediansurvival of ∼3 years after diagnosis (Demedts et al., 2001). IPF is asubcategory under the general disease category of interstitial lungdisease (ILD). Common traits of ILDs are the presence ofinflammation and pulmonary fibrosis. IPF patients have a poorprognosis, with 44% mortality rate for IPF patients expected within5 years, compared with the mortality rates in two other ILDs, ILD-associated connective tissue disease (33%) and pulmonarysarcoidosis (2%) (Demedts et al., 2001).

Like many fibrotic disorders, IPF is characterized by enhanceddeposition and remodelling of the extracellular matrix (ECM).Although data are emerging about the altered composition of theECM in IPF (Booth et al., 2012; Estany et al., 2014; Kuhn et al., 1989;Kuhn and McDonald, 1991; Ward and Hunninghake, 1998), less isknown about the structural changes that occur in collagens in thisdisease. Collagens undergo extensive post-translational modifications,including formation of cross-links between fibres that stabilize themolecules and generate a strong collagen fibril, which provides tensilestrength. Aberrant cross-linking is known to contribute to increasedtissue stiffness, a pathological characteristic of IPF (Clarke et al., 2013).

Fibrillar collagen structures (predominantly collagens I and III)can be visualized and quantified using second harmonic generation(SHG) microscopy (Abraham et al., 2012; Tjin et al., 2014). SHGimaging has been used to analyse ECM structures in biologicalsystems in vitro, ex vivo and in vivo (Brown et al., 2003; Caetano-Lopes et al., 2010; Cox et al., 2003; Guo et al., 1999; Nadiarnykhet al., 2007; Sivaguru et al., 2010; Suzuki et al., 2012; Williamset al., 2001) and can semi-quantitatively measure the ratio ofmature organized collagen to immature disorganized collagenfibrils (Williams et al., 2005). This technique, therefore, enablescomparisons of collagen rearrangement in tissues, analysis ofcollagen structure (Caetano-Lopes et al., 2010; Nadiarnykh et al.,2007; Sivaguru et al., 2010; Suzuki et al., 2012) and evenmeasurement of the thickness of the fibres (Chu et al., 2007). Wehave recently reported the robustness of this technique to quantifyfibrillar collagen I remodelling in airway tissues in chronicobstructive pulmonary disease (Tjin et al., 2014).

Maturity of collagen fibres and thus the mechanical properties ofthe fibres are dependent on the level of intermolecular cross-linking.Received 11 April 2017; Accepted 16 August 2017

1Respiratory Cellular and Molecular Biology Group, Woolcock Institute of MedicalResearch, Glebe, New South Wales 2037, Australia. 2Central Clinical School,Faculty of Medicine, The University of Sydney, Sydney, New South Wales 2006,Australia. 3Australian Centre for Microscopy and Microanalysis, The University ofSydney, Sydney, New South Wales 2006, Australia. 4Stem Cell Regulation Unit, St.Vincent’s Institute of Medical Research, Victoria 3065, Australia. 5Pulmonary &Critical Care Medicine, Department of Internal Medicine, University of Michigan,Ann Arbor, MI 48109, USA. 6University of Groningen, University Medical CenterGroningen, Department of Pulmonary Diseases, Groningen, 9713 GZ, TheNetherlands. 7University of Groningen, University Medical Center Groningen,Department of Pathology & Medical Biology, Experimental Pulmonology andInflammation Research, Groningen, 9713 GZ, The Netherlands. 8University ofGroningen, University Medical Center Groningen, GRIAC (Groningen ResearchInstitute for Asthma and COPD), Groningen, 9713 GZ, The Netherlands.9Department of Physiology & Biomedical Engineering, College of Medicine, MayoClinic, Rochester, MN 55905, USA. 10Department of Tissue Pathology andDiagnostic Oncology, Royal Prince Alfred Hospital, Sydney, New South Wales2050, Australia. 11Discipline of Pharmacology, The University of Sydney, Sydney,New South Wales 2006, Australia.

*Authors for correspondence (gtjin@svi.edu.au; j.k.burgess@umcg.nl)

G.T., 0000-0002-5125-2801; A.F., 0000-0003-1740-3538; D.S., 0000-0002-6570-3212; D.J.T., 0000-0002-5115-9025; A.M., 0000-0003-2268-9737; E.P.W.K.,0000-0002-2741-0793; J.K.B., 0000-0001-9868-9966

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1301

© 2017. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

Although the intermolecular fibrillar organization and certaincross-linking processes can happen spontaneously, enzymes suchas lysyl oxidases (LOs) are necessary for further cross-linking. TheLO family of enzymes consists of five known paralogues: lysyloxidase (LOX) and LOX like 1-4 (LOXL1-4). During normaldevelopment, the expression of LOs is tightly regulated, withaberrant expression and activity being associated with pathologicalmanifestation in various diseases, including cancer (Erler et al.,2006; Payne et al., 2005) and, as recently discovered, IPF (Aumilleret al., 2017; Barry-Hamilton et al., 2010).Collagen cross-linking contributes to the stiffness of the ECM of

tissues and is tightly regulated in normal tissue homeostasis.However, in diseased tissue, aberrant cross-linking could lead tothe development of a pathological microenvironment. The stiffness ofthe ECM influences the behaviour and function of cells within itsvicinity, including an increase in fibroblast proliferation andcontraction (Balestrini et al., 2012; Marinkovic et al., 2013). Inaddition, a stiffer ECM leads to an increase in latent transforminggrowth factor-β (TGF-β) activation (Burgess et al., 2016; Shi et al.,2011; Wipff et al., 2007). In IPF, the lung tissue has a stiffness of16.52±2.25 kPa, as opposed to 1.96±0.13 kPa in healthy lung tissue(Booth et al., 2012). A stiffer matrix also promotes IPF-likephenotypes, such as enhanced differentiation, proliferation andresistance to apoptosis in non-IPF fibroblasts (Liu et al., 2010;Marinkovic et al., 2012; Paszek et al., 2005; Wipff et al., 2007).Despite the availability of two recently approved drugs for slowingthe progress of IPF [nintedanib (Mazzei et al., 2015; Richeldi et al.,2014a,b) and pirfenidone (King et al., 2014; Valeyre et al., 2014)],their associated incidence of side effects and modest clinical benefitwarrants further investigation into possible new therapeutic directionsfor IPF (Moodley et al., 2015). Selective inhibition of LOXL2 activitywas recently investigated as a therapeutic approach for IPF [Barry-Hamilton et al., 2010; ClinicalTrials.gov NCT01769196 (Raghuet al., 2017)]. Although this trial was terminated early because of lackof efficacy, there is still potential for the development of effective LOinhibitors for targeting fibrosis. To enhance understanding of themechanisms by which LOX and LOXL enzyme inhibition might bebeneficial in IPF, this study aimed to investigate fibrillar collagenstructural remodelling in IPF lung tissues and to determine theexpression profile of LOs in IPF. It also investigated the effectivenessof pan-LO and selective LOXL2 inhibition on fibrillar collagenstructural remodelling and tissue stiffness in the context of IPF.

RESULTSIncreased fibrillar collagen maturity/organization in IPFlung tissuesTo investigate the collagen structural remodelling in IPF, SHGmicroscopy was performed on lung tissue sections from non-IPFand IPF subjects. The SHG forward (F) signal is predominantlyfrom mature/organized fibrillar collagen, whereas the backward (B)signal is from immature/disorganized fibrillar collagen. Thus,increased F/B area corresponds to increased amounts of mature/immature collagen, whereas increased F/B intensity ratioscorrespond to an increased degree of collagen maturity/thicknessand vice versa (Tjin et al., 2014). Quantification of SHG imagesshowed that IPF lung parenchymal tissue had greater levels offibrillar collagen maturity/organization compared with the non-IPFtissues (Fig. 1A,B). By contrast, the collagen structural organizationwas not different between IPF and non-IPF ILD tissues (Fig. S1).

Lysyl oxidase genes were differentially expressed in IPF andhealthy control lung tissuesFibrillar collagen maturity is catalysed by the LO family ofenzymes. To investigate the levels of LO enzymes in lung tissues,lung tissue gene expression profiles of IPF and healthy controlswere analysed from two publically available data sets (Table 1).LOXL1 gene expression was increased in IPF lung tissue comparedwith healthy controls in two independent studies [false discoveryrate (FDR) adjusted P-value <0.25], whereas the LOXL2 geneexpression was increased in IPF lung tissue in a single data set. Nodifferences were found in LOX gene expression levels.

Lysyl oxidase protein levels were differentially expressedbetween IPF and non-diseased lung tissuesTo confirm the findings from the gene expression study, tissuesections from the same samples used for the examination ofcollagen structural organization were examined for LOX, LOXL1and LOXL2 protein levels. Quantification of the captured images ofthe stained lung tissue sections showed that LOX percentage surfacearea and density were decreased in IPF compared with non-IPF,whereas the percentage surface area and density of LOXL1 wasincreased in IPF compared with non-IPF tissues (Table 2;representative images in Fig. 2A). In agreement with prior studies(Aumiller et al., 2017; Barry-Hamilton et al., 2010), LOXL2 densitywas also increased in IPF compared with non-IPF tissues. By

Fig. 1. Increased level of collagen organization/maturity in IPF lung parenchyma. (A) Representative SHG images of lung parenchyma from non-IPF (n=7;open circles) and IPF (n=8; filled diamonds) subjects (average of three or four samples per subject; yellow indicates backward immature/disorganizedcollagen; cyan indicatesmature/organized collagen; scale bar: 100 µm). Brightness and contrast have been enhanced for display purposes only andwere appliedequally for all images. (B) Image analysis quantification of SHG intensity forward/backward signal ratio (intensity F/B; each point represents the mean ofthree or four lung parenchymal samples per subject); a higher ratio indicates higher proportion of organized/mature fibrillar collagen. Data were analysed withStudent’s two-tailed parametric unpaired t-test. Data are presented as means±s.d. Data were obtained from one experimental replicate per subject. *P<0.05.

1302

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

contrast, only LOXL1, but not LOX or LOXL2, density increased inIPF compared with non-IPF ILD (Table S1). In vitro, primaryhuman lung parenchymal fibroblasts from IPF patients producedsignificantly greater levels of LOXL2 but not LOX or LOXL1 in thecell lysate compared with non-IPF controls (Fig. S2A,B).

Expression of LOs was correlated with collagen remodellingin lung tissuesTo begin to understand the relationship between LOs and collagenstructural arrangement in lung tissues, we used correlation analysis.The data from both non-IPF and IPF subjects were combined in thecorrelation studies to examine the whole spectrum of collagenremodelling in both cohorts. There was a positive correlationbetween both LOXL1 (Table 3; Fig. S3) and LOXL2 (Table 3) withcollagen remodelling (SHG F/B ratios), whereas there was anegative correlation between LOX and SHG forward to backwardsignal (F/B) ratios (Table 3). LOXL1 positively correlated withLOXL2, whereas LOX negatively correlated with both LOXL1 andLOXL2 (Table 3).

Lysyl oxidase family members were differentially expressedin IPFThe ratio of LOX/LOXL1 (Fig. 2B) and LOX/LOXL2 (Fig. S4)densities and LOX/LOXL1 percentage surface areas weresignificantly decreased in IPF lung tissue samples compared withnon-IPF subjects (Fig. 2B; Fig. S4). The ratio of LOX/LOXL1

percentage surface area inversely correlated with collagen maturity/organization (F/B; Fig. 2C). There was no significant difference inthe ratios of LO family enzymes in the tissues of IPF and non-IPFILD subjects (Fig. S5; P=NS).

Collagen organization and LOX enzymes interrelatedifferently in IPFGiven that LOX, LOXL1 and LOXL2 were differentially expressedin IPF compared with non-IPF (Fig. 2B,C) and the expression ofthese enzymes also correlated strongly with collagen remodelling(Table 3), it was important to understand the interrelationshipbetween these parameters in IPF. Visualization of the three-dimensional (3D) association among these variables showed spatialseparations between the variables in IPF compared with non-IPF(Fig. 2D). We next performed factor analysis to investigate thenumber of relationship groups (factors) accounting for the structureof these associations. A minimum of two groups were required toexplain the associations within the data set (Fig. 2E), and Table S2shows the strength of the contribution of the variables [LOX,LOXL1, LOXL2, immunohistochemistry (IHC) data and SHG data]to the two relationship groups. The strengths of contributions of thevariables were used to determine the weighted average values foreach relationship group (Fig. 2F,G). The values for relationshipGroups 1 and 2were significantly different between non-IPFand IPF,indicating a difference in the relationships between the variables(LOX, LOXL1, LOXL2 IHC data and SHG data) in disease.

Inhibition of LO activity reduced TGF-β-induced collagenremodelling in collagen I hydrogelsTo confirm the role of LO enzymes in fibrillar collagen remodelling,3D in vitro cell culture experiments using inhibitors of LO activitywere performed. β-APN is a non-selective inhibitor of LO activity,whereas Compound A is a specific inhibitor of LOXL2 activity(Schilter et al., 2014). Pan-inhibition of LO activity, but notinhibition of LOXL2-specific activity, reduced TGF-β-inducedcollagen remodelling in collagen hydrogel fibroblast cultures(Fig. 3). TGF-β increased the deposition of mature fibrillarcollagen by non-IPF fibroblasts but not IPF fibroblasts, whichwas alleviated by the pan-LO inhibitor β-APN (Fig. 3D), but TGF-βdid not affect collagen fibre maturity/thickness (Fig. 3E). Bycontrast, Compound A did not alleviate the increased deposition ofmature collagen compared with immature collagen, from bothfibroblast groups, induced by TGF-β, but on the contrary, increasedit (Fig. 3F). Compound A did reduce the thickness of the fibresgenerated by the IPF fibroblasts in the presence and absence ofTGF-β stimulation (Fig. 3G).

Inhibition of LO activity reduced TGF-β-induced collagenremodelling in reseeded decellularized matricesTo model the in vivo cell ECM interactions more closely, we useddecellularized human IPF and non-IPF lung scaffolds. There weresignificantly more primary human lung fibroblasts remaining in thesupernatant after reseeding of the IPFmatrices compared with the non-IPF matrices (Fig. 4A), potentially indicating higher cell attachment tothe non-IPF matrix. This was supported by histological staining of thereseeded matrices, which showed that fewer cells were detectable inthe IPF matrix post-reseeding, compared with the non-IPF matrix(Fig. 4B). Given the sparsity of cellular attachment in the IPF matrices,these matrices were not examined further.

The non-IPF cells increased the maturity/thickness of thecollagen fibres in the non-IPF matrix in the presence of TGF-β(Fig. 4C,D) but did not increase the amount of mature collagen

Table 1. LOX expression levels in lung tissue

Gene ID

Fold change(non-IPFversus IPF) P-value FDR Data set

ILMN_11693 1.04 8.05×10−1 0.97 GSE48149LOX 215446_s_at 0.87 3.71×10−1 0.59 GSE24206

204298_s_at 0.92 6.26×10−1 0.79 GSE24206213640_s_at 0.96 6.57×10−1 0.81 GSE24206

LOXL1 ILMN_7655 1.39 5.21×10−3 0.11 GSE48149203570_at 2.51 4.81×10−4 0.02 GSE24206

LOXL2 ILMN_22633 1.03 7.17×10−1 0.96 GSE48149202998_s_at 1.57 2.11×10−2 0.13 GSE24206228808_s_at 0.92 2.42×10−1 0.47 GSE24206202999_s_at 0.93 3.57×10−1 0.58 GSE24206202997_s_at 1.04 6.64×10−1 0.81 GSE24206

FDR, false discovery rate.

Table 2. Lysyl oxidase enzymes are differentially expressed in IPF lungtissues

Non-IPF IPFImage analysis Mean±s.d. Mean±s.d. P-value

LOX % Surface area 95.58±3.23 87.17±7.12 ↓ 0.009*†

Density 211.10±10.07 193.20±9.90 ↓ 0.016*LOXL1 % Surface area 66.54±13.20 79.86±10.28 ↑ 0.033*

Density 171.80±16.56 188.90±12.14 ↑ 0.027*LOXL2 % Surface area 50.84±23.76 66.86±9.85 0.094

Density 134.30±14.05 153.00±5.76 ↑ 0.003*

Quantification of immunohistochemical staining of the lung tissues usingimage analysis on whole tissue. Data are presented as the percentage of thesurface area (% Surface area) stained with the protein of interest and density ofstaining within the stained surface areas, comparing non-IPF (n=10) and IPF(n=8) subjects. Data were analysed with Student’s two-tailed parametricunpaired t-test unless specified. Data were obtained from one experimentalreplicate per subject for each antibody. ↓, significant decrease comparedwith non-IPF; ↑, significant increase compared with non-IPF. *P<0.05.†Mann–Whitney U-test.

1303

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

compared with immature collagen (Fig. S6). The pan-inhibitor ofLOs, β-APN, returned TGF-β-induced collagen remodelling tobaseline. In addition, the LOXL2-specific inhibitor blocked TGF-β-induced remodelling of fibrillar collagen; remodelling levels in thepresence of either inhibitor without TGF-β were not differentcompared with control (Fig. 4D). Primary IPF lung fibroblastsseeded into decellularized normal matrices did not remodel thefibrillar collagen in the presence of TGF-β (Fig. 4D).

Inhibition of LO activity reduced TGF-β-induced stiffening inreseeded decellularized matricesTo confirm the functional significance of TGF-β-induced collagenremodelling in the decellularized matrices reseeded with non-IPFcells, tissue stiffness (elastic modulus) was measured by atomicforce microscopy (AFM; micro-indentation). TGF-β induced asignificant increase in tissue stiffness that was inhibited by the pan-inhibitor of LOs, β-APN (Fig. 5). Individual stiffness measurements

Fig. 2. Collagen organization and lysyloxidase family of enzymes aredifferentially expressed and interrelatedifferently in IPF. (A) Representative IHCimages of lysyl oxidase enzymes in thelungs. Images are representative of lungtissues from non-IPF (n=10) and IPF (n=8)subjects (brown, protein of interest; blue,nucleus; pink, cytoplasm; scale bar:200 µm). (B) The ratio of LOX and LOXL1percentage surface area obtained byimage analysis of IHC-stained lung tissuesof non-IPF (n=10; open circles) comparedwith IPF (n=8; filled diamonds) subjects;data were analysed with Student’s two-tailed parametric unpaired t-test.(C) Correlation of the ratio of LOX andLOXL1 with F/B ratio obtained throughimage analysis of SHGmicroscopy imagesof collagen structure in non-IPF (n=7) andIPF (n=8) subjects (P<0.0001; R2=0.7527;average of three or four samples persubject). (D) 3D scatter plot of collagenorganization (intensity F/B, i.e. SHGintensity forward/backward signal ratio),LOXL1 and LOXL2 expression (opencircles, non-IPF, n=7; filled circles, IPF,n=8). (E) Scree plot showing the number ofrelationship groups needed to explain theassociations within measured variables innon-IPF and IPF tissues (eigenvaluethreshold≥1). (F,G) Weighted averagevalues of the variables contributing to eachrelationship group; comparing non-IPF withIPF subjects (average of three or foursamples per subject); data were analysedwith Student’s two-tailed parametricunpaired t-test. Data are presented asmeans±s.d. Data were obtained from oneexperimental replicate per subject.*P<0.05, **P<0.01, ***P<0.005.

1304

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

for each patient segregated by treatment can be found in Fig. S7, andrepresentative force curves used to measure stiffness can be found inFig. S8C.

DISCUSSIONAlthough increased collagen deposition is well appreciated infibrotic lung tissues, this study identifies, for the first time,alterations in the fibrillar organization of the collagen fibres inIPF lung tissue. In addition, this study reports that LOXL1 andLOXL2 gene and protein levels are increased in the lungs of IPFcompared with non-IPF subjects, with a positive correlationbetween increased protein expression and increased fibrillarcollagen organization.These findings show the complex interrelationship between LOs

and collagen structure and provide further support for focusing onLO inhibition as a potential therapeutic approach in IPF. Indeed,LOXL2 inhibition prevented and even reversed bleomycin-inducedpulmonary fibrosis in mice (Barry-Hamilton et al., 2010). Inaddition, LOX and LOXL2 were also found to be elevated in liverfibrosis (Barry-Hamilton et al., 2010; Vadasz et al., 2005) andwere associated with increased deposition of collagen aroundthe hepatocytes (Vadasz et al., 2005). Although the recentclinical trial targeting LOXL2 inhibition (ClinicalTrials.govNCT01769196) was halted early because of a lack of efficacy(Raghu et al., 2017), the present study suggests that targetingLOXL1 in IPF might be as important as targeting LOXL2.In the present study, IPF alveoli had higher collagen maturity/

thickness (increased F/B SHG) compared with non-diseasedsubjects. It can be postulated that the increased collagen fibrematurity/thickness contributes to the increased stiffnesscharacteristic of IPF lungs (Booth et al., 2012). This is supportedby the increased stiffnesses measured in this study in thedecellularized lung scaffolds reseeded with fibroblasts exposed toTGF-β. In IPF there is augmented TGF-β (Broekelmann et al.,1991), which induces IPF fibroblasts to produce enhanced ECMproteins (Westergren-Thorsson et al., 2004), thereby creating apositive feedback loop. In fact, collagen cross-linking mightaccount for organ stiffening in a variety of fibrotic diseases. Theresults of the present study are also concordant with those ofKottmann et al. (2015), who demonstrated an altered F/B SHG

signal in IPF compared with non-diseased lungs. Interestingly, thestudy by Kottmann et al. (2015) showed a decrease in F/B SHGbetween healthy and IPF lungs; this difference might be attributableto different sampling and analysis protocols in comparison to thepresent study. For example, in the study by Kottmann et al. (2015),the majority of the F SHG signal (numerator in the F/B SHG signalratio) from healthy subjects originated from muscular arteries,whereas the present study excluded such regions from analysis. Thatsaid, neither study identified differences in collagen remodellingbetween IPF and other fibrotic interstitial lung diseases.

A potential role of other LO family members, not only LOXL2, incollagen remodelling in IPF was also highlighted in the presentstudy. A significant increase in LOXL1 gene and protein expressionin IPF lungs, compared with non-IPF tissues, was observed, with apositive correlation of protein levels with collagen organization. Bycontrast, LOX gene expression was unchanged, whereas the LOXprotein level was decreased in IPF lung, highlighting differential LOfamily expression in IPF lung tissue, which is similar to thatreported during tumour progression and metastasis (Erler et al.,2006; Payne et al., 2005). In concordancewith our findings, a recentstudy by Aumiller et al. (2017) also reported increased LOXL1 andLOXL2 protein levels in IPF lungs compared with healthy lungs. Bycontrast, however, they also reported an increased LOX protein levelin IPF (Aumiller et al., 2017). These differential results might beattributable to the different analysis approaches. Aumiller et al.(2017) scored the intensity of LO staining in discrete structures ofthe lung, whereas in our study we used digitized image analysis tomeasure the staining intensity of the whole lung tissue section whilstcontrolling for the increased tissue mass in IPF lungs.

Of the LO family enzymes, only the LOXL1 protein level wasincreased in IPF compared with non-IPF ILD lungs, indicating thatchanges in LOXL1 might be specific to IPF rather than to theprocess of fibrosis itself. To our knowledge, this study is the first toshow a difference in the interrelationships of the organization ofcollagen, LOX, LOXL1 and LOXL2 enzymes between the lungs ofIPF and non-IPF subjects. This is particularly important, because aPhase II clinical trial using simtuzumab, a LOXL2-specific inhibitor[ClinicalTrials.gov NCT01769196 (Raghu et al., 2017)] wasrecently terminated early owing to a lack of efficacy. The datapresented in the present study support the continued development of

Table 3. Expression of lysyl oxidase enzymes correlates with the maturity/organization of the fibrillar collagen and other lysyl oxidase enzymes inlung tissues

Correlation matrix

F/B ratio LOX LOXL1 LOXL2

Area Intensity % Surface area Density % Surface area Density % Surface area Density

F/B ratio Area 1.000 0.896<0.001*

−0.4200.060

−0.5120.025*

0.767<0.001*

0.866<0.001*

0.7340.001*

0.834<0.001*

Intensity 1.000 −0.3830.079

−0.5060.027*

0.821<0.001*

0.866<0.001*

0.6340.006*

0.6870.002*

LOX % Surface area 1.000 0.882<0.001*

0.0470.434

−0.1000.361

−0.3820.080

−0.6710.003*

Density 1.000 −0.0710.401

−0.2540.180

−0.3670.089

−0.7300.001*

LOXL1 % Surface area 1.000 0.911<0.001*

0.7070.002*

0.5320.021*

Density 1.000 0.6230.007*

0.6450.005*

LOXL2 % Surface area 1.000 0.7540.001*

Density 1.000

Correlation matrix of IHC staining of the lung tissues from both non-IPF (n=7) and IPF (n=8) subject groups and second harmonic imaging F/B ratios (upper value,R2 values; lower values, P-values). Data were obtained from one experimental replicate per subject for each antibody. F/B ratio, SHG forward/backwardsignal ratio. *P<0.05.

1305

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

new therapeutic strategies targeting LOs for IPF, especially focusingon exploration of the potential of LOXL1-specific inhibitors orcombination approaches.To understand the role of LOs in the remodelling of collagen

fibres in IPF, two 3D in vitro culture models were used to mimic thecomplexity of the matrix in vivo. Interestingly, the decellularizedIPF matrices supported lower primary fibroblast infiltration and

viability compared with non-IPF matrices. Although prior works(Booth et al., 2012; Parker et al., 2014) are in line with the presentdata that decellularized IPF matrices influence cellular phenotype,previous evaluations had not examined differences in cellattachment between non-IPF and IPF matrices. Thus, theseobservations highlight another difference between IPF and non-IPF lung matrices that should be investigated as a contributor to the

Fig. 3. Inhibition of lysyl oxidase activity reducesTGF-β-induced collagen remodelling. Collagen Ihydrogels seeded with primary human lungparenchymal fibroblasts (250,000 cells ml−1) werestimulated with TGF-β (10 ng ml−1) to induce a fibroticphenotype in the presence or absence of the pan-lysyloxidase inhibitor, β-APN (100 µM), or the LOXL2-specific inhibitor, Compound A (Cmpd A; 300 nM).Samples were treated for 7 days. At the end of thetreatment period, the samples were formalin fixed andparaffin embedded. (A) Haematoxylin and Eosin stainof fibroblasts in collagen gel (blue, nucleus; pink,cytoplasm and general tissue stain; scale bar: 100 µm)at day 7. (B) SHG image of collagen gel with fibroblastsat day 7 (cyan, forward mature/organized collagen;magenta, cell autofluorescence; scale bar=100 µm).(C) Representative SHG images of collagen gels withprimary human lung parenchymal fibroblasts treatedwith or without TGF-β (10 ng ml−1) in the presence orabsence of Compound A (300 nM) (yellow, backwardimmature/disorganized collagen; cyan, forwardmature/organized collagen; scale bar: 20 µm).(D-G) Quantification of SHG microscopy on collagen Ihydrogels embedded with fibroblasts from non-IPF(n=5; open circles) and IPF (n=5; filled diamonds)subjects treatedwith pan-lysyl oxidase inhibitor (β-APN;D,E) or LOXL2-specific inhibitor (Compound A; F,G).Data are presented as a percentage of the respectiveno-cell control (means±s.d.). (D,F) SHG forward/backward signal surface area ratio (F/B area); higherratio indicates higher amount of mature compared withimmature collagens. (E,G) SHG forward/backwardsignal intensity ratio (F/B intensity); higher ratioindicates greater maturity/thickness level of fibrillarcollagen fibres. Statistical analysis was by two-wayANOVA with matching and multiple comparisons withTukey’s correction. Data are presented as means±s.d.Datawere obtained from one experimental replicate persubject. Brightness and contrast have been enhancedfor display purposes only and were applied equally forall images. *Significant difference compared with BSAcontrol; δsignificant difference compared with BSA+Compound A control; ϕsignificant difference comparedwith TGF-β alone; and #significant difference betweengroups.

1306

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

pathogenesis of fibrosis. These findings further highlight thechallenge ahead in finding solutions for repairing the disruptedECM in IPF.In collagen gels, TGF-β induced fibroblast production of mature/

thick collagen fibres, but did not affect the overall maturity/thickness of the fibres. By contrast, TGF-β did not significantly alterthe amount of mature collagen fibres in decellularized human lungmatrices, whereas overall maturity/thickness was increased. Thisapparent paradox might simply be attributable to the modelcompositions, in that collagen hydrogels lack the complex ECMcomponents found in decellularized matrices, components thatmight influence cellular responsiveness. Alternatively, theartificiality of the collagen hydrogel, where collagen fibres are not

in a native structural configuration, leads cells to respond bydepositing and remodelling collagen fibres to attain a nativeenvironment. By contrast, decellularized matrices are composed ofcollagen fibres and ECM already in a native configuration, therebydriving cellular driven ECM remodelling rather than production.This latter conclusion is supported by the presented data (Figs 4 and5) showing collagen gel F/B intensity >150% of the no cell controleven in unstimulated conditions, whereas decellularized matriceswere recorded at ∼100% in the same conditions. The differences inthe presence of artificial matrix (collagen hydrogel) and nativematrix (decellularized matrix) support previous findings that theECM contributes a very important regulatory control on cellularbehaviour (Booth et al., 2012; Parker et al., 2014).

Fig. 4. Inhibition of lysyl oxidase activityreduces TGF-β-induced collagenremodelling. Decellularized non-IPFmatrices reseeded with primary human lungparenchymal fibroblasts (1×106 cells permatrix) were stimulated with TGF-β(10 ng ml−1) to induce a fibrotic phenotype inthe presence or absence of the pan-lysyloxidase inhibitor, β-APN (100 µM), or theLOXL2-specific inhibitor, Compound A (CmpdA; 300 nM). Cells were seeded in the matrixfor 24 h before being transferred into separatetissue culture wells with treatment media for7 days. Samples were collected at the end ofthe treatment period and formalin fixed andparaffin embedded. (A) Manual cell counts ofcells remaining in supernatant after reseedingin matrices [n=9; combined data from 4 non-IPF and 5 IPF fibroblasts (500,000 cells ml−1;2 ml per matrix) into matrices from one non-IPF and one IPF subject]. (B) Masson’sTrichrome stain of fibroblasts reseeded intodecellularized matrices for 7 days (blue,collagen; pink, cell cytoplasm; scale bar:100 µm). (C) Representative SHG images ofcollagen gels with fibroblasts treated with orwithout TGF-β (10 ng ml−1) in the presence ofor absence of Compound A (300 nM) (yellow,backward immature/disorganized collagen;cyan, forward mature/organized collagen;scale bar: 100 µm). (D) Quantification of SHGmicroscopy on decellularized non-IPFmatrices reseeded for 7 days with fibroblastsfrom control (n=5; open circles) and IPF (n=5;filled diamonds) subjects treated with β-APN(100 µM) or Compound A (300 nM) in thepresence or absence of TGF-β (10 ng ml−1).Data are presented as a percentage of therespective no-cell control (means±s.d.); SHGforward/backward signal intensity ratio (F/Bintensity); higher ratio indicates greatermaturity/thickness level of fibrillar collagenfibres. Statistical analysis was by two-wayANOVA with matching and multiplecomparisons with Tukey’s correction. Dataare presented as means±s.d. Data wereobtained from one experimental replicate persubject. Brightness and contrast have beenenhanced for display purposes only and wereapplied equally for all images. *Significantdifference compared with BSA control; and#significant difference between groups.

1307

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

Pan-inhibition of LO activity blocked TGF-β-induced collagenremodelling in both collagen hydrogels and decellularized matrixsystems. By contrast, LOXL2-specific inhibition reduced collagenfibril maturity/thickness only in IPF fibroblasts and in decellularizedmatrices. As this effect was not seen with the pan-inhibitor of LOs, itis likely that this effect was independent of LOXL2. However, priordata demonstrate that fibroblasts grown in two-dimensional (2D)culture express a 50 kDa proteolytically processed form of LOXL2(Hollosi et al., 2009), rather than the intact 95 kDa LOXL2 enzymethat is inhibited by Compound A. Although the functional activity ofthe LOXL2 50 kDa isoform is unknown, it seems unlikely to accountfor the present observations, because Compound Awas able to blockother TGF-β-induced effects in 2D and 3D culture. Additionally, pan-inhibition of LO activity blocked TGF-β-induced tissue stiffeningin the decellularized matrix system (Fig. 5). Both collagen fibrilmaturity/thickness and tissue stiffness were increased by TGF-βtreatment, with both being interrupted in the presence of β-APN.These data indicate that LO enzymes regulate the TGF-β-inducedincrease in collagen fibril maturity/thickness that results in increasedtissue stiffness. The pro-fibrotic features of stiff matrices contribute tothe TGF-β-induced fibrosis with potential feedback mechanismsthrough the promotion of latent TGF-β activation (Wipff et al., 2007),thus contributing to the fibrotic process in IPF.It is worth noting that the stiffness measured in the reseeded

decellularized matrices (mean 40-60 kPa) was higher than thevalues measured in native human lung parenchymal tissue (1.34±0.36 kPa; Liu et al., 2010) and IPF tissue (16.52±2.25 kPa; Boothet al., 2012). It has been reported in the literature that thedecellularization method affects the stiffness of acellular tissue(Melo et al., 2014) and, more generally, that decellularized tissue isstiffer than normal lung tissue (Giménez et al., 2017), although theorigins of these differences in decellularized versus normal tissuestiffness remain poorly understood. The reseeded decellularizedmatrices did not form alveolar structures in the time span of thisstudy and were densely packed (Fig. 4B), possibly as a result of lackof inflation of the tissues, which is not possible within the presentexperimental set-up. This tissue structure might more closely mimicscar tissue or remodelled IPF lung tissue (characterized byprogressive scarring of the lungs) rather than non-diseased lungtissues. In such a case, this would strengthen the significance

of reseeded decellularized matrices in the study of IPF tissueremodelling. It should also be noted that β-APN might have non-cell-induced effects on the stiffness, as the stiffness appears lower inβ-APN-treated non-reseeded matrices than those without β-APNtreatment (∼40 versus ∼60 kPa, respectively; Fig. S7), but thepresent experiment was not designed to address this hypothesis.

Overall, this study highlights a potential mechanism forpreventing ongoing fibrotic processes, with LO inhibitionpreventing lung fibrogenesis in IPF by blocking fibrillar collagenorganization and subsequent tissue stiffening. This, in turn, mightleave deposited collagen more susceptible to proteolyticdegradation and clearance, although further research is needed toinvestigate this possibility. The present findings also highlight therole of multiple members of the LO family in regulating collagencross-linking and the disrupted balance of LOs in the IPF lung.These findings indicate that selective inhibition of LO enzymes,such as LOXL1, either alone or in combination with LOXL2inhibition, might be a potential therapeutic strategy for targetingfibrosis in IPF.

MATERIALS AND METHODSHuman tissuesHuman lungs were procured with written and informed consent from thepatients or next of kin from non-transplantable donors and IPF patientsundergoing lung transplantation as previously described (Booth et al.,2012). Approval for all experiments with human lung tissue was providedby the Ethics Review Committees of the South West Sydney Area HealthService, St Vincent’s Hospital Sydney, Royal Prince Alfred Hospital(RPAH), and the University of Sydney Human Research Ethics Committee.As tissues were de-identified, the University of Michigan InstitutionalReview Board deemed this work exempt from oversight. Primary humanlung fibroblast samples used as non-diseased controls (non-IPF) were fromthe macroscopically normal regions of lungs from patients with non-smallcell carcinoma (NSCCa). Human lung tissue samples were classified bypathologist diagnosis. Patient demographics are described in Table S3.

Tissue and cell isolationTissue isolation was performed on fresh explanted lung tissues. Human lungparenchymal tissues were isolated and cut into pieces no bigger than 1 cm3.These pieces were fixed in 4% formaldehyde (Fronine; Thermo FisherScientific, Waltham, MA, USA). After formaldehyde fixation, isolated lungtissues were embedded in paraffin.

Fig. 5. Inhibition of lysyl oxidase activity reduces TGF-β-induced stiffening of reseeded decellularized matrices.Decellularized non-IPF matrices reseeded with primary humanlung parenchymal fibroblasts (106 cells per matrix) werestimulated with TGF-β (10 ng ml−1) in the presence or absenceof the pan-lysyl oxidase inhibitor, β-APN (100 µM). Cells wereseeded in the matrix for 24 h before being transferred intoseparate tissue culture wells with treatment media for 7 days.Samples were collected at the end of the treatment period andfresh frozen in OCT compound (10 µm slices). Box and whiskersplot of the stiffness (elastic modulus). Stiffness was measuredwith AFM, and >100 measurements spread over three areaswere taken per sample. Each n is a single AFM measurement.Data were obtained from one experimental replicate per subjectper treatment. Statistical analysis was by the Kruskal–Wallis testwith Dunn’s correction. **P<0.01, ****P<0.0001.

1308

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

Primary human lung fibroblasts were isolated as previously described(Krimmer et al., 2012; White et al., 2003) and were grown in growth media[Dulbecco’s modified Eagle’s medium (DMEM) low glucose (Gibco,Thermo Fisher Scientific) with 0.025 M HEPES (Gibco), 0.375% sodiumhydrogen carbonate (Ajax Finechem, Thermo Fisher Scientific), 10% (v/v)FBS (JRH Biosciences, Brooklyn, Victoria, Australia) and 1% (v/v)antibiotics (Invivogen, San Diego, CA, USA), pH 7.1-7.2]. All fibroblastcell lines were tested for mycoplasma contamination after cell isolation andculture. For experiments, mycoplasma-free fibroblasts were used betweenthe third and ninth passages.

In vitro cell culturePrimary human lung fibroblasts were grown in growth media [DMEM lowglucose (Gibco) with 0.025 M HEPES (Gibco) and 0.375% sodiumhydrogen carbonate (Ajax Finechem, Thermo Fisher Scientific)] incubatedin a humidified CO2 incubator (5% CO2 in air) at 37°C. Growth media wereaspirated from cells every 4 days and replaced with fresh growth media untilconfluence.

Protein extraction from primary human lung fibroblastsWhole-cell lysates were prepared using protein extraction buffer [20 mMTris (Amresco, Dallas, TX, USA) pH 7.4 with 150 mM sodium chloride(PanReac Applichem, Darmstadt, Germany), 1 mM EGTA (Sigma, StLouis, MO, USA), 1 mM EDTA (Amresco), 1 mM sodium fluoride(NaF; Sigma), 20 mM sodium pyrophosphate (Sigma), 2 mM sodiumorthovanadate (Na3VO4; Sigma), 1% (v/v) Triton X-100 (Sigma), 10% (v/v)glycerol (Sigma), 0.1% (w/v) SDS (Amresco), 0.5% (w/v) sodiumdeoxycholate (Sigma), 1 mM PMSF (Life Technologies, Solon, OH,USA), 1% (v/v) Protease Inhibitor Cocktail Set III (Calbiochem, San Diego,CA, USA)] on ice, scraped into a tube, and centrifuged at 150×g for 5 min.The supernatants were collected and stored at −80°C for furtherexperiments. Total sample protein was measured using a standard BCAassay (Sigma).

SHG microscopySHG imaging and analysis were performed on paraffin-embedded,formalin-fixed lung tissue sections as previously published (Tjin et al.,2014).

Thirty-micrometre-thick samples were used for SHG analysis in humantissue samples (cohort 1, non-IPF=7 versus IPF=8; and cohort 2, non-IPFILD=8 versus IPF=26) and reseeded decellularized human lung tissuesamples. Five-micrometre-thick tissue samples were used for the SHG incollagen gel samples, as the softness of the sample prevented the use ofthicker samples because of the unavoidable compression of these samplesduring processing. Confocal images were taken at ×63 magnification at16-bit resolution for both forward and backward propagated SHG signals,with identical detector configurations for all images taken. Laser power wasmeasured at the objective before the start of each experiment performed ondifferent days, and laser power settings were adjusted so that all experimentsuse the same laser power (25 mW).

Owing to the heterogeneous nature of the lung tissues, a 10×10 grid wasapplied on the human tissue samples using the tile function of the Leica LASAF software (Leica, Wetzlar, Germany). Three regions for imaging werethen selected using a random number generator. A 5×5 grid was used ondecellularized tissue samples because of the smaller size of the samplescompared with the size of human lung tissue samples. Collagen gel sampleswere more homogeneous, and three regions were randomly selectedfor SHG.

In the figures, adjustments of brightness and contrast as indicated in thefigure legends were applied uniformly for every set of images and were usedfor display purposes only.

SHG image analysisImage analysis was performed using Fiji (Schindelin et al., 2012) withimages imported from the LAS AF software (Leica). The stack wasdeliberately oversampled by 20% in each direction to ensure that the entiretyof the tissue section was imaged. The confocal 3D imagewas analysed using

the histogram to determine the mean (µ) and s.d. (σ) of the signal for theimage. A threshold of µ+σ was then applied to the image to excludebackground emissions as previously published (Abraham et al., 2012).

The pixel area (the number of pixels with intensity above threshold) andpixel density (average signal intensity per pixel with intensity abovethreshold) was measured for every slice in the image using batch processing.The total signal intensity (total intensity for all pixels with intensity abovethreshold) was determined as the product of the pixel area and density. Theaverage values for these measurements were calculated for every image inthe stack. The above process was repeated for both forward and backwardpropagated data to generate area, intensity and density data for both forwardand backward propagated signals. The ratio of forward to backward signalwas then calculated for area, intensity and density measurements.

Lysyl oxidase gene expression in lung tissuesTwo publically available microarray data sets from lung tissue samples ofIPF patients and healthy controls were analysed (GSE48149, non-IPF=9versus IPF=13; and GSE24206, non-IPF=6 versus IPF=10). Microarrayanalysis was conducted using R software version 3.02, using theBioconductor-limma package, and normalized using Robust Multi-arrayAverage (RMA). A Benjamini–Hochberg procedure (false discovery rate)was undertaken to adjust for multiple testing.

Lysyl oxidase protein expression in lung tissuesImmunohistochemistry was performed using a protocol adapted from Faizet al. (2013)with antibodies against LOX (Abcam, Cambridge, UK; ab31238;www.abcam.com/lox-antibody-ab31238.html, accessed 2 May 2016),LOXL1 (Novus Biologicals, Littleton, CO, USA; NBP1-82827; www.novusbio.com/LOXL1-Antibody_NBP1-82827.html, accessed 2 May 2016)and LOXL2 (Novus Biologicals; NBP1-32954; www.novusbio.com/Lysyl-Oxidase-Homolog-2-LOXL2-Antibody_NBP1-32954.html, accessed 2 May2016). Primary antibodies were as follows: LOX (1 µg ml−1), LOXL1(0.1 µg ml−1) or LOXL2 (1 µg ml−1) and isotype control antibody (sameconcentration as the respective primary antibodies). Following washing,samples were treated with EnVision+ anti-rabbit horseradish peroxidase(HRP)-conjugated secondary antibody (K4003, Dako, Santa Clara, CA,USA). After a second wash step, colour development was detected with3′,3′-diaminobenzidine (DAB; Dako), and the tissues were thencounterstained in ready-to-use Mayer’s Haematoxylin solution (Sigma) for5 min. Sections were then dehydrated, mounted and coverslipped. Allsections for each primary antigen were stained in the same staining run.

IHC image analysisImages were captured using aWide-field FL and TLmicroscope ZEISS AxioScan.Z1 Slide Scanner (Zeiss, Oberkochen, Germany). The immunostainedtissuewas scanned at ×20 magnification, and a whole brightfield tissue imagewas stitched together from the individual serial images using ZEN Software(Zeiss). All images for each set of immunostaining were captured on the sameday with the same microscope settings to ensure uniformity and accuracy ofquantified image analysis. Fiji ImageJ (Schindelin et al., 2012) was used toquantify the density and distribution of staining. Colour deconvolution(Ruifrok and Johnston, 2001) vectors in ImageJ were optimized to ensureaccurate separation of Eosin and DAB. Overlaid thresholded images of Eosinand DAB were used to measure total tissue surface area. The image analysiscalculated the number of pixels above the threshold within the image (‘area’)and the average intensity of the pixels above the threshold (‘averageintensity’). Macros were used to batch process the images.

Data were represented as ‘percentage tissue surface area’ (DAB area/totaltissue surface area) and ‘density’ (average intensity calculated for the wholetissue).

Lysyl oxidase protein expression in cell lysatesWestern immunoblot was performed using antibodies against LOX (Abcam;ab31238; 1:2000), LOXL1 (Novus Biologicals; NBP1-82827; 1:2000) andLOXL2 (Novus Biologicals; NBP1-32954; 1:2000) using a protocoladapted from Burgess et al. (2003). Cell lysates were run on 10%polyacrylamide gels and transferred to polyvinyledene difluoride

1309

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

membranes (Merck Millipore, Darmstadt, Germany) and protein blockedfor 1 h in 5% skimmed milk. After washing, samples were treated withHRP-conjugated secondary antibody (Dako). After a second wash step,detection was performed using chemiluminsecent reagent ImmobilonWestern (Merck Millipore) and bands were analysed using a Kodakimaging system and CarestreamMI software (Molecular Bioimaging, Bend,OR, USA). Membranes were stripped and reprobed for GAPDH (MerckMillipore; 1:5000 dilution) as above as a loading control.

Adjustments of brightness and contrast as indicated in the figure legendswere applied uniformly for every set of images and used for displaypurposes only.

Antifibrotic potential of LO activity inhibitors (collagen gel3D model)Collagen gel solution was prepared for a final collagen gel concentration of2 mg ml−1. Briefly, 10% (v/v) of 10× PBS and 10% (v/v) 0.1 M NaOHwereadded to stock rat-tail collagen I (BD Biosciences, Franklin Lakes, NJ, USA)gel solution and mixed while cooled on ice. DMEM media [DMEM lowglucose (Gibco) with 0.025 MHEPES (Gibco) and 0.375% sodium hydrogencarbonate (Ajax Finechem)] containing resuspended cells was then added athigher cell concentration for dilution in the collagen gel solution (finalconcentration of 2.5×105 cells ml−1 in the collagen gels). Collagen gel mixedwith cells (n=5 non-IPF, n=5 IPF cell lines and n=1 no-cell control) waspipetted into a 48-well plate and allowed to set in a humidified CO2 incubatorat 37°C for 2 h. Once the gel had set, the gels were then released from thewells by gently running a sterile syringe needle around between the gel andthe wall of the wells. Treatment media [Set 1: 0.1% bovine serum albumin(BSA) ±10 ng ml−1 TGF-β ±0.1 mM β-APN; Set 2: 0.01% DMSO +0.1%BSA ±10 ng ml−1 TGF-β ±300 nM Compound A (Pharmaxis, Sydney,Australia)] of the same volume as the gel were pipetted gently on top of thegels in the wells. The collagen gels were collected at day 7, fixed in 4%paraformaldehyde solution and paraffin embedded.

Decellularization of human lung tissuesDecellularized tissues from a non-diseased subject and an IPF subject wereprepared as previously described (Booth et al., 2012). Essentially, lungtissues were sequentially incubated in sterile water, deoxycholic acid(Sigma), sodium chloride (PanReac) and DNAse (Sigma) to clear cellularand nuclear materials from the tissues. Tissues were washed three times insterile PBS between each step. The resultant acellular tissue was biopsiedusing a 12-mm punch and then sliced into 1-mm slices on a vibratome. Theacellular tissues slices were sterilized in peracetic acid (Sigma)/ethanol andrinsed extensively in sterile PBS. Slices were stored in sterile PBS at 4°Cuntil use.

Antifibrotic potential of LO activity inhibitors (decellularizedmatrix 3D model)Decellularized lung slices from both non-diseased and IPF subjects werewashed three times in sterile DMEM solution and were subsequently addedto a suspension of 5.0×105 fibroblasts ml−1 in growth media at a ratio of 1matrix:2 ml of cell solution (n=5 non-IPF, n=5 IPF cell lines and n=1 no-cellcontrol) up to a maximum of 25 ml of solution in a 50 ml Falcon tube. TheFalcon tubes containing the matrices and cell suspension were mounted on asuspension mixer (Selby, ThermoFisher Scientific, Waltham, MA, USA)and constantly rotated at low speed (approximately one rotation every 20 s)at 37°C for 24 h. The reseeded ECMs were placed in separate wells of a48-well plate with fresh media [0.01% DMSO +0.1% BSA ±10 ng ml−1

TGF-β (R&D Systems, MN, USA) ±300 nM Compound A or 0.1 mMβ-APN (Sigma)] and incubated at 37°C and 5% CO2. The reseededdecellularized matrices were collected at day 7; half were fixed in 4%paraformaldehyde solution for 24 h and paraffin embedded, whereas theother half were fresh frozen in optimal cutting temperature (OCT)compound and stored at −80°C.

Tissue stiffness measurements using atomic force microscopyAFMmeasurements were performed on decellularized human lung matriceswith or without reseeded non-IPF cells. Matrices were embedded in optimal

OCT compound, and 10-µm-thick tissue slices were cryosectioned at−21°Cand mounted on poly-L-lysine-coated glass slides. PBS solution was addedon the tissue slice to avoid drying.

Parenchymal tissue areas were identified under an optical microscope (×200magnification; Olympus, Tokyo, Japan) and mechanically analysed with aBioScope Catalyst AFM (Bruker, Billerica, MA, USA). Microindentationswere performed using a 2.5-µm-radius sphere-tipped probe (Novascan, Ames,IA, USA) with a spring constant determined at ∼100 pN nm−1 by the thermalfluctuation method (Thundat et al., 1994). For each sample, three randomlyselected areas from three non-consecutive tissue slices were analysed in PBS atroom temperature. Force curves were acquired with MIRO 2.0 (NanoScope9.1; Bruker) at an indentation rate of 20 μm s−1 and a ramp size of 10 μm atdifferent points along the parenchymal tissue.More than 100 force curves wereperformed per sample (∼35 per area; Fig. S8).

The Young’s modulus, E, was determined by fitting of the force curve bythe Hertz sphere model (Dimitriadis et al., 2002; Hertz, 1881) usingNanoScope Analysis software (Bruker) and considering Poisson’s ratio of0.4 (Butler et al., 1986).

Statistical analysisData were entered and sorted using Microsoft Excel (Microsoft, Redmond,WA,USA). Sample sizes for comparing IPF versus non-IPFwere determinedbased on a previous study (Tjin et al., 2014). Statistical analysis wasperformed using GraphPad Prism Version 6 Software (GraphPad, La Jolla,CA, USA). Data were tested for normal distribution with the D’Agostino andPearson omnibus normality test and analysed with Student’s t-test (normallydistributed) or the Mann–Whitney U-test (non-normally distributed) asappropriate. Kruskal–Wallis and two-way ANOVA statistical analyses wereperformed where appropriate. Factor analysis was performed using SPSSsoftware (IBM, Armonk, NY, USA) using ‘principal axis factoring’analysing ‘correlation matrix’ and extracting ‘2’ factors from the data set.Promax rotation was used for the factor analysis, using a regression methodand excluding cases listwise where there were missing data points.

AcknowledgementsThe authors would like to acknowledge the subjects of the study, thecardiopulmonary transplant team and pathologists at St Vincent’s Hospital and theteam at the Department of Tissue Pathology and Diagnostic Oncology, Royal PrinceAlfred Hospital. The authors acknowledge the facilities and the scientific andtechnical assistance of the Australian Microscopy &Microanalysis Research Facilityat the Australian Centre for Microscopy & Microanalysis at the University of Sydneyand the University of Sydney Histopathology laboratory. Dr Wolfgang Jarolimek andDr Heidi Schilter from Pharmaxis kindly provided Compound A. Mr Mario D’Souza, aresearch statistician employed by Sydney Local Health District Clinical ResearchCentre, provided assistance with the statistical analysis.

Competing interestsThe authors declare no competing or financial interests.

Author contributions metadataConceptualization: G.T., E.P.W.K., J.K.B.; Methodology: G.T., E.P.W.K.; Software:G.T.; Validation: G.T., A.F., D.S., D.J.T.; Formal analysis: G.T., E.S.W., A.F., D.S.,D.J.T., A.M., E.P.W.K., J.K.B.; Investigation: G.T., A.F., D.S.; Resources: G.T.,E.S.W., D.J.T., A.M., E.P.W.K., J.K.B.; Data curation: G.T., J.K.B.; Writing - originaldraft: G.T.; Writing - review& editing: G.T., E.S.W., A.F., D.S., D.J.T., A.M., E.P.W.K.,J.K.B.; Visualization: G.T., A.F., D.S.; Supervision: E.P.W.K., J.K.B.; Projectadministration: G.T., E.S.W., D.J.T., E.P.W.K., J.K.B.; Funding acquisition: J.K.B.

FundingJ.K.B. was supported by a National Health and Medical Research Council AustraliaFellowship (1032695) and a Rosalind Franklin Fellowship, co-funded byRijksuniversiteit Groningen, Universitair Medisch Centrum Groningen and theEuropean Union. A.F. was funded by a RESPIRE2 fellowship from the EuropeanRespiratory Society. D.J.T., D.S. and E.S.W. were funded by the National Institutesof Health (HL092961, HL133320 and U01HL111016; UL1TR00043 andUL1TR002240 to E.S.W.).

Data availabilityTwo publically available microarray data sets from lung tissue samples of IPF andhealthy controls, GSE48149 and GSE24206, were used in this study.

1310

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.030114.supplemental

ReferencesAbraham, T., Kayra, D., McManus, B. and Scott, A. (2012). Quantitativeassessment of forward and backward second harmonic three dimensionalimages of collagen Type I matrix remodeling in a stimulated cellular environment.J. Struct. Biol. 180, 17-25.

Aumiller, V., Strobel, B., Romeike, M., Schuler, M., Stierstorfer, B. E. and Kreuz,S. (2017). Comparative analysis of lysyl oxidase (like) family members inpulmonary fibrosis. Sci. Rep. 7, 149.

Balestrini, J. L., Chaudhry, S., Sarrazy, V., Koehler, A. and Hinz, B. (2012). Themechanical memory of lung myofibroblasts. Integr. Biol. 4, 410-421.

Barry-Hamilton, V., Spangler, R., Marshall, D., McCauley, S., Rodriguez, H. M.,Oyasu, M., Mikels, A., Vaysberg, M., Ghermazien, H., Wai, C. et al. (2010).Allosteric inhibition of lysyl oxidase-like-2 impedes the development of apathologic microenvironment. Nat. Med. 16, 1009-1017.

Booth, A. J., Hadley, R., Cornett, A. M., Dreffs, A. A., Matthes, S. A., Tsui, J. L.,Weiss, K., Horowitz, J. C., Fiore, V. F., Barker, T. H. et al. (2012). Acellularnormal and fibrotic human lung matrices as a culture system for in vitroinvestigation. Am. J. Respir. Crit. Care. Med. 186, 866-876.

Broekelmann, T. J., Limper, A. H., Colby, T. V. and McDonald, J. A. (1991).Transforming growth factor beta 1 is present at sites of extracellular matrix geneexpression in human pulmonary fibrosis. Proc. Natl. Acad. Sci. USA 88,6642-6646.

Brown, E., McKee, T., diTomaso, E., Pluen, A., Seed, B., Boucher, Y. and Jain,R. K. (2003). Dynamic imaging of collagen and its modulation in tumors in vivousing second-harmonic generation. Nat. Med. 9, 796-801.

Burgess, J. K., Johnson, P. R. A., Ge, Q., Au, W. W., Poniris, M. H., McParland,B. E., King, G., Roth, M. and Black, J. L. (2003). Expression of connective tissuegrowth factor in asthmatic airway smooth muscle cells. Am. J. Respir. Crit. Care.Med. 167, 71-77.

Burgess, J. K., Mauad, T., Tjin, G., Karlsson, J. C. andWestergren-Thorsson, G.(2016). The extracellular matrix - the under-recognized element in lung disease?J. Pathol. 240, 397-409.

Butler, J. P., Nakamura, M., Sasaki, H., Sasaki, T. and Takishima, T. (1986).Poissons’ ratio of lung parenchyma and parenchymal interaction with bronchi.Jpn. J. Physiol. 36, 91-106.

Caetano-Lopes, J., Nery, A. M., Canhao, H., Duarte, J., Cascao, R., Rodrigues,A., Perpetuo, I. P., Abdulghani, S., Amaral, P. M., Sakaguchi, S. et al. (2010).Chronic arthritis leads to disturbances in the bone collagen network. Arthritis Res.Ther. 12, R9.

Chu, S.-W., Tai, S.-P., Chan, M.-C., Sun, C.-K., Hsiao, I.-C., Lin, C.-H., Chen, Y.-C.and Lin, B.-L. (2007). Thickness dependence of optical second harmonicgeneration in collagen fibrils. Opt. Express 15, 12005-12010.

Clarke, D. L., Carruthers, A. M., Mustelin, T. and Murray, L. A. (2013). Matrixregulation of idiopathic pulmonary fibrosis: the role of enzymes. FibrogenesisTissue Repair 6, 20.

Cox, G., Kable, E., Jones, A., Fraser, I., Manconi, F. and Gorrell, M. D. (2003).3-dimensional imaging of collagen using second harmonic generation. J. Struct.Biol. 141, 53-62.

Demedts, M., Wells, A. U., Anto, J. M., Costabel, U., Hubbard, R., Cullinan, P.,Slabbynck, H., Rizzato, G., Poletti, V., Verbeken, E. K. et al. (2001). Interstitiallung diseases: an epidemiological overview. Eur. Respir. J. Suppl. 32, 2s-16s.

Dimitriadis, E. K., Horkay, F., Maresca, J., Kachar, B. and Chadwick, R. S.(2002). Determination of elastic moduli of thin layers of soft material using theatomic force microscope. Biophys. J. 82, 2798-2810.

Erler, J. T., Bennewith, K. L., Nicolau, M., Dornhofer, N., Kong, C., Le, Q.-T., Chi,J.-T. A., Jeffrey, S. S. and Giaccia, A. J. (2006). Lysyl oxidase is essential forhypoxia-induced metastasis. Nature 440, 1222-1226.

Estany, S., Vicens-Zygmunt, V., Llatjos, R., Montes, A., Penın, R., Escobar, I.,Xaubet, A., Santos, S., Manresa, F., Dorca, J. et al. (2014). Lung fibrotictenascin-C upregulation is associated with other extracellular matrix proteins andinduced by TGFbeta1. BMC Pulm. Med. 14, 120.

Faiz, A., Tjin, G., Harkness, L., Weckmann, M., Bao, S., Black, J. L., Oliver,B. G. G. and Burgess, J. K. (2013). The expression and activity of cathepsins D,H and K in asthmatic airways. PLoS ONE 8, e57245.

Gimenez, A., Uriarte, J. J., Vieyra, J., Navajas, D. and Alcaraz, J. (2017). Elasticproperties of hydrogels and decellularized tissue sections used inmechanobiology studies probed by atomic force microscopy. Microsc. Res.Tech. 80, 85-96.

Guo, Y., Savage, H. E., Liu, F., Schantz, S. P., Ho, P. P. and Alfano, R. R. (1999).Subsurface tumor progression investigated by noninvasive optical secondharmonic tomography. Proc. Natl. Acad. Sci. USA 96, 10854-10856.

Hertz, H. (1881). Üeber die Beruhrung fester elastischer Korper. J. Reine Angew.Math. 92, 156-171.

Hollosi, P., Yakushiji, J. K., Fong, K. S. K., Csiszar, K. and Fong, S. F. T. (2009).Lysyl oxidase-like 2 promotes migration in noninvasive breast cancer cells but notin normal breast epithelial cells. Int. J. Cancer 125, 318-327.

King, T. E., Jr, Bradford, W. Z., Castro-Bernardini, S., Fagan, E. A., Glaspole, I.,Glassberg,M. K., Gorina, E., Hopkins, P.M., Kardatzke, D., Lancaster, L. et al.(2014). A phase 3 trial of pirfenidone in patients with idiopathic pulmonary fibrosis.N. Engl. J. Med. 370, 2083-2092.

Kottmann, R. M., Sharp, J., Owens, K., Salzman, P., Xiao, G.-Q., Phipps, R. P.,Sime, P. J., Brown, E. B. and Perry, S. W. (2015). Second harmonic generationmicroscopy reveals altered collagen microstructure in usual interstitial pneumoniaversus healthy lung. Respir. Res. 16, 61.

Krimmer, D. I., Burgess, J. K.,Wooi, T. K., Black, J. L. andOliver, B. G. G. (2012).Matrix proteins from smoke-exposed fibroblasts are pro-proliferative.Am. J. Respir. Cell Mol. Biol. 46, 34-39.

Kuhn, C. and McDonald, J. A. (1991). The roles of the myofibroblast in idiopathicpulmonary fibrosis. Ultrastructural and immunohistochemical features of sites ofactive extracellular matrix synthesis. Am. J. Pathol. 138, 1257-1265.

Kuhn, C., III, Boldt, J., King, T. E., Jr, Crouch, E., Vartio, T. and McDonald, J. A.(1989). An immunohistochemical study of architectural remodeling andconnective tissue synthesis in pulmonary fibrosis. Am. Rev. Respir. Dis. 140,1693-1703.

Liu, F., Mih, J. D., Shea, B. S., Kho, A. T., Sharif, A. S., Tager, A. M. andTschumperlin, D. J. (2010). Feedback amplification of fibrosis through matrixstiffening and COX-2 suppression. J. Cell Biol. 190, 693-706.

Marinkovic, A., Mih, J. D., Park, J. A., Liu, F. and Tschumperlin, D. J. (2012).Improved throughput traction microscopy reveals pivotal role for matrix stiffness infibroblast contractility and TGF-beta responsiveness. Am. J. Physiol. Lung Cell.Mol. Physiol. 303, L169-L180.

Marinkovic, A., Liu, F. and Tschumperlin, D. J. (2013). Matrices of physiologicstiffness potently inactivate idiopathic pulmonary fibrosis fibroblasts.Am. J. Respir. Cell Mol. Biol. 48, 422-430.

Mazzei, M. E., Richeldi, L. and Collard, H. R. (2015). Nintedanib in the treatment ofidiopathic pulmonary fibrosis. Ther Adv Respir Dis. 9, 121-129.

Melo, E., Garreta, E., Luque, T., Cortiella, J., Nichols, J., Navajas, D. and Farre,R. (2014). Effects of the decellularization method on the local stiffness of acellularlungs. Tissue Eng. Part C Methods 20, 412-422.

Moodley, Y., Corte, T., Richeldi, L. and King, T. E., Jr. (2015). Do all patients withidiopathic pulmonary fibrosis warrant a trial of therapeutic intervention? A pro-conperspective. Respirology 20, 389-394.

Nadiarnykh, O., Plotnikov, S., Mohler, W. A., Kalajzic, I., Redford-Badwal, D.and Campagnola, P. J. (2007). Second harmonic generation imagingmicroscopy studies of osteogenesis imperfecta. J. Biomed. Opt. 12, 051805.

Parker, M. W., Rossi, D., Peterson, M., Smith, K., Sikstrom, K., White, E. S.,Connett, J. E., Henke, C. A., Larsson, O. and Bitterman, P. B. (2014). Fibroticextracellular matrix activates a profibrotic positive feedback loop. J. Clin. Invest.124, 1622-1635.

Paszek, M. J., Zahir, N., Johnson, K. R., Lakins, J. N., Rozenberg, G. I., Gefen, A.,Reinhart-King, C. A., Margulies, S. S., Dembo, M., Boettiger, D. et al. (2005).Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241-254.

Payne, S. L., Fogelgren, B., Hess, A. R., Seftor, E. A., Wiley, E. L., Fong, S. F. T.,Csiszar, K., Hendrix, M. J. C. and Kirschmann, D. A. (2005). Lysyl oxidaseregulates breast cancer cell migration and adhesion through a hydrogen peroxide-mediated mechanism. Cancer Res. 65, 11429-11436.

Raghu, G., Collard, H. R., Egan, J. J., Martinez, F. J., Behr, J., Brown, K. K.,Colby, T. V., Cordier, J.-F., Flaherty, K. R., Lasky, J. A. et al. (2011). An officialATS/ERS/JRS/ALAT statement: idiopathic pulmonary fibrosis: evidence-basedguidelines for diagnosis and management. Am. J. Respir. Crit. Care. Med. 183,788-824.

Raghu, G., Brown, K. K., Collard, H. R., Cottin, V., Gibson, K. F., Kaner, R. J.,Lederer, D. J., Martinez, F. J., Noble, P. W., Song, J. W. et al. (2017). Efficacy ofsimtuzumab versus placebo in patients with idiopathic pulmonary fibrosis: arandomised, double-blind, controlled, phase 2 trial. Lancet Respir. Med. 5, 22-32.

Richeldi, L., Cottin, V., Flaherty, K. R., Kolb, M., Inoue, Y., Raghu, G., Taniguchi,H., Hansell, D. M., Nicholson, A. G., Le Maulf, F. et al. (2014a). Design of theINPULSIS™ trials: two phase 3 trials of nintedanib in patients with idiopathicpulmonary fibrosis. Respir. Med. 108, 1023-1030.

Richeldi, L., du Bois, R. M., Raghu, G., Azuma, A., Brown, K. K., Costabel, U.,Cottin, V., Flaherty, K. R., Hansell, D. M., Inoue, Y. et al. (2014b). Efficacy andsafety of nintedanib in idiopathic pulmonary fibrosis. N. Engl. J. Med. 370,2071-2082.

Ruifrok, A. C. and Johnston, D. A. (2001). Quantification of histochemical stainingby color deconvolution. Anal. Quant. Cytol. Histol. 23, 291-299.

Schilter, H., Buson, A., Cock, T.-A., Deodhar, M., Findlay, A. D., Foot, J. S.,Jarnicki, A., Moses, J., Turner, C. I., Yow, T. T. et al. (2014). Inhibition of LOXL2:Pharmaxis’ small molecule approach to treat fibrosis. InC30. LateBreakingAbstractsin Disease Treatment and Clinical Outcomes, pp. A6669-A6669, American ThoracicSociety 2014 International Conference. New York: ATS Journals.

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch,T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B. et al. (2012). Fiji: anopen-source platform for biological-image analysis. Nat. Methods 9, 676-682.

1311

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms

Shi, M., Zhu, J., Wang, R., Chen, X., Mi, L., Walz, T. and Springer, T. A. (2011).Latent TGF-beta structure and activation. Nature 474, 343-349.

Sivaguru, M., Durgam, S., Ambekar, R., Luedtke, D., Fried, G., Stewart, A. andToussaint, K. C., Jr. (2010). Quantitative analysis of collagen fiber organization ininjured tendons using Fourier transform-second harmonic generation imaging.Opt. Express 18, 24983-24993.

Suzuki, M., Kayra, D., Elliott, W. M., Hogg, J. C. and Abraham, T. (2012). Secondharmonic generation microscopy differentiates collagen type I and type III indiseased lung tissues. Proc. SPIE 8226, Multiphoton Microscopy in theBiomedical Sciences X11, 82263F.

Thundat, T., Warmack, R. J., Chen, G. Y. and Allison, D. P. (1994). Thermal andambient-induced deflections of scanning force microscope cantilevers. Appl.Phys. Lett. 64, 2894-2896.

Tjin, G., Xu, P., Kable, S. H., Kable, E. P. W. and Burgess, J. K. (2014).Quantification of collagen I in airway tissues using second harmonic generation.J. Biomed. Opt. 19, 036005.

Vadasz, Z., Kessler, O., Akiri, G., Gengrinovitch, S., Kagan, H. M., Baruch, Y.,Izhak, O. B. and Neufeld, G. (2005). Abnormal deposition of collagen aroundhepatocytes in Wilson’s disease is associated with hepatocyte specific expressionof lysyl oxidase and lysyl oxidase like protein-2. J. Hepatol. 43, 499-507.

Valeyre, D., Albera, C., Bradford, W. Z., Costabel, U., King, T. E., Jr, Leff, J. A.,Noble, P. W., Sahn, S. A. and du Bois, R. M. (2014). Comprehensive

assessment of the long-term safety of pirfenidone in patients with idiopathicpulmonary fibrosis. Respirology 19, 740-747.

Ward, P. A. and Hunninghake, G. W. (1998). Lung inflammation and fibrosis.Am. J. Respir. Crit. Care. Med. 157, S123-S129.

Westergren-Thorsson, G., Sime, P., Jordana, M., Gauldie, J., Sarnstrand, B.and Malmstrom, A. (2004). Lung fibroblast clones from normal and fibroticsubjects differ in hyaluronan and decorin production and rate of proliferation.Int. J. Biochem. Cell Biol. 36, 1573-1584.

White, E. S., Thannickal, V. J., Carskadon, S. L., Dickie, E. G., Livant, D. L.,Markwart, S., Toews, G. B. and Arenberg, D. A. (2003). Integrin alpha4beta1regulates migration across basement membranes by lung fibroblasts: a role forphosphatase and tensin homologue deleted on chromosome 10. Am. J. Respir.Crit. Care. Med. 168, 436-442.

Williams, R. M., Zipfel, W. R. and Webb, W. W. (2001). Multiphoton microscopy inbiological research. Curr. Opin. Chem. Biol. 5, 603-608.

Williams, R. M., Zipfel, W. R. and Webb, W. W. (2005). Interpreting second-harmonic generation images of collagen I fibrils. Biophys. J. 88, 1377-1386.

Wipff, P. J., Rifkin, D. B., Meister, J. J. and Hinz, B. (2007). Myofibroblastcontraction activates latent TGF-beta1 from the extracellular matrix. J. Cell Biol.179, 1311-1323.

1312

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1301-1312 doi:10.1242/dmm.030114

Disea

seModels&Mechan

isms